Monadic Logging & Evaluation in Haskell

[mkanta]

This is the Haskell version of safe evaluation and division counting plus a combination of both using
monad transformers. Mostly untested, but it illustrates the monad manipulations involved.

Also some updates concerning the January meeting, some questions answered in more detail.

[vgrocha]

• Added some comments and questions
• Suggested changes to MonadEval.hs so it can be loaded in ghci

[vgrocha]

It's an interesting idea to have logDivs wrap around the evalSim, but it looks like this implementation will apply f every time it recurses, leading to unnecessary evaluations. Was this intended?

[vgrocha]

ah, I see. You wrap it in an Identity Monad when you want to turn a simple value into a monadic value to be elsewhere

[vgrocha]

Suggested change

	import Control.Monad.Writer
	import Control.Monad
	import Control.Monad.Identity
	import Control.Monad.Writer
	import Data.Monoid

[vgrocha]

Suggested change

	logDivUSafe = logDivU evalSafe
	logDivUSafe = logDivsU evalSafe

[vgrocha]

Suggested change

	logDivUSim = logDivU (Identity . evalSim)
	logDivUSim = logDivsU (Identity . evalSim)

[mkanta]

Thanks for your comments. I'm terribly busy with taxes at the moment, so it'll probably take me until sometime next week-end to tackle that. Cheers, Matthias

…

On Sun, 10 Mar 2019 at 15:39, Vinicius ***@***.***> wrote: ***@***.**** commented on this pull request. - Added some comments and questions - Suggested changes to MonadEval.hs so it can be loaded in ghci ------------------------------ In 2019-02-28-Monad-structure/MonadEval.hs <#9 (comment)>: > +evalSafe (Cons x) = return x + +-- |For logging, the monad to use is @writer w@ because the logging +-- value of type @w@ is accumulated automatically when chaining through +-- monadic values. That's why @w@ is required to be an instance of @monoid@, +-- which allows the accumulation of values with the monoidal composition. +-- Whilst @State@ could be used as well, the accumulation would have to be +-- done manually. +-- For counting divisions, a @writer (Sum Int) Int@ monad is +-- called for because @sum Int@ is the monoid with @0@ as neutral element +-- and @+@ as operation. +logDivs :: (Term -> Int) -> Term -> Writer (Sum Int) Int +logDivs f arg@(Div (x,y)) = logDivs f x >>= logSecond + where + logSecond n = logDivs f y >>= combineWith n + combineWith nm den = writer (f $ Div (Cons nm, Cons den),1) It's an interesting idea to have logDivs wrap around the evalSim, but it looks like this implementation will apply f every time it recurses, leading to unnecessary evaluations. Was this intended? ------------------------------ In 2019-02-28-Monad-structure/MonadEval.hs <#9 (comment)>: > + where + logSecond n = logDivsU f y >>= combineWith n + combineWith nm den = WriterT (f (Div (Cons nm, Cons den)) >>= packTup) + packTup x = return (x,1) +logDivsU f arg = WriterT $ f arg >>= \x -> return (x,0) + +-- | This can be used with @evalsafe@ +logDivUSafe :: Term -> WriterT (Sum Int) (Either String) Int +logDivUSafe = logDivU evalSafe + +-- | But also with @evalSim@, however @logDivUSim=logDivU evalSim@ does not +-- work because the value returned by @evalSim@ is not monadic. This is +-- where the @Identity@ monad comes in handy because it allows to +-- pack the value of @evalSim@ for use with @logDivsU@: +logDivUSim :: Term -> WriterT (Sum Int) Identity Int +logDivUSim = logDivU (Identity . evalSim) ah, I see. You wrap it in an Identity Monad when you want to turn a simple value into a monadic value to be elsewhere ------------------------------ In 2019-02-28-Monad-structure/MonadEval.hs <#9 (comment)>: > @@ -0,0 +1,100 @@ +module MonadEval + where + +import Control.Monad.Writer ⬇️ Suggested change -import Control.Monad.Writer +import Control.Monad +import Control.Monad.Identity +import Control.Monad.Writer +import Data.Monoid ------------------------------ In 2019-02-28-Monad-structure/MonadEval.hs <#9 (comment)>: > +-- which in our case would be @m ~ Either ***@***.*** Note how little the +-- code has to change from @logDivs@ to its universal version @logDivsU@: +-- All the @writer@ functions have to be replaced by the @WriterT@ constructors +-- with some minor adjustments to account for the different argument +-- requirements. +logDivsU :: Monad m => (Term -> m Int) -> Term -> WriterT (Sum Int) m Int +logDivsU f arg@(Div (x,y)) = logDivsU f x >>= logSecond + where + logSecond n = logDivsU f y >>= combineWith n + combineWith nm den = WriterT (f (Div (Cons nm, Cons den)) >>= packTup) + packTup x = return (x,1) +logDivsU f arg = WriterT $ f arg >>= \x -> return (x,0) + +-- | This can be used with @evalsafe@ +logDivUSafe :: Term -> WriterT (Sum Int) (Either String) Int +logDivUSafe = logDivU evalSafe ⬇️ Suggested change -logDivUSafe = logDivU evalSafe +logDivUSafe = logDivsU evalSafe ------------------------------ In 2019-02-28-Monad-structure/MonadEval.hs <#9 (comment)>: > + where + logSecond n = logDivsU f y >>= combineWith n + combineWith nm den = WriterT (f (Div (Cons nm, Cons den)) >>= packTup) + packTup x = return (x,1) +logDivsU f arg = WriterT $ f arg >>= \x -> return (x,0) + +-- | This can be used with @evalsafe@ +logDivUSafe :: Term -> WriterT (Sum Int) (Either String) Int +logDivUSafe = logDivU evalSafe + +-- | But also with @evalSim@, however @logDivUSim=logDivU evalSim@ does not +-- work because the value returned by @evalSim@ is not monadic. This is +-- where the @Identity@ monad comes in handy because it allows to +-- pack the value of @evalSim@ for use with @logDivsU@: +logDivUSim :: Term -> WriterT (Sum Int) Identity Int +logDivUSim = logDivU (Identity . evalSim) ⬇️ Suggested change -logDivUSim = logDivU (Identity . evalSim) +logDivUSim = logDivsU (Identity . evalSim) — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#9 (review)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AlnP_AX3ivEM9bV-VLj3lYV-VRWmj5bJks5vVVFvgaJpZM4bbYpJ> .